论文部分内容阅读
目的结合肺癌危险因素研究中变量的筛选过程，探讨在涉及较多自变量的大型多元回归分析中，变量间多重共线性的诊断和处理方法。方法首先将经单因素分析筛选的变量进行相关分析，得出相关系数矩阵Ｒ的特征值，用主成分分析法判定自变量间是否存在多重共线性以及存在几个多重共线性关系。然后将这些自变量进行正交旋转，取得旋转后公因子所对应的自变量及其多重共线性关系，结合专业知识和以往研究的经验加以去除。结果将去除多重共线性的自变量引入多元回归模型，即可取得比较满意的结果。结论在大型多元回归分析中用上述方法进行多重共线性的诊断和处理是可行的。
Objective To explore the process of screening for variables in the study of risk factors for lung cancer and to explore the methods of diagnosis and treatment of multicollinearity among variables in large multivariate regression analysis involving more independent variables. Methods Firstly, the correlation coefficients of the correlation coefficient matrix R were obtained by correlation analysis of the variables selected by single factor analysis. Principal component analysis was used to determine whether there was multicollinearity among the independent variables and there were several multicollinearity relationships. Then, these independent variables are orthogonally rotated, and the independent variables corresponding to the common factors after rotation and their multiple collinearity relationships are obtained, which are removed by combining professional knowledge and past research experience. As a result, multiple independent collinearity independent variables were introduced into the multiple regression model to obtain satisfactory results. Conclusion It is feasible to diagnose and treat multicollinearity using the above method in large multivariate regression analysis.